Programmers who published open-source code on GitHub sued GitHub, Microsoft, and various OpenAI entities, alleging that GitHub Copilot and Codex—AI tools trained on publicly available code from GitHub—reproduce portions of their code without attribution. These programmers claimed that the AI’s omission of copyright management information (CMI), such as attribution and license terms required by open-source licenses, violated the Digital Millennium Copyright Act (DMCA), specifically 17 U.S.C. § 1202(b). Plaintiffs alleged that Copilot’s outputs sometimes consist of verbatim or near-verbatim reproductions of their code, but the AI-generated outputs do not include the original CMI.

The United States District Court for the Northern District of California reviewed the case and dismissed the DMCA claims under Rule 12(b)(6, first with leave to amend and then with prejudice, concluding that plaintiffs failed to allege that Copilot’s outputs were “identical” to their code and that only identical copies from which CMI had been removed could support a DMCA claim. The court allowed breach of contract claims to proceed. It certified the DMCA dismissal for interlocutory appeal under 28 U.S.C. § 1292(b), noting the issue of whether § 1202(b) imposes an identicality requirement.

The United States Court of Appeals for the Ninth Circuit affirmed the district court’s dismissal. The court held that plaintiffs had Article III standing due to a plausible risk of injury. However, it determined that under their “output” theory, Copilot and Codex do not “remove or alter” CMI from copies of existing protected works; instead, they generate new works that never contained CMI. The court declined to consider the plaintiffs’ “input” theory as it was forfeited. The main holding is that generating new works without CMI does not violate § 1202(b) of the DMCA. View “DOE V. GITHUB, INC.” on Justia Law